Abstract
The focus of the present thesis is to develop the simulation capability to analyze the performance of the Chalcogenide phase change memory devices. The simulations for the thermal behavior, the electrical properties and the crystallization of the Phase-Change Memory in RESET and SET states have been conducted using the energy equation, the charge conservation law, and nucleation theories. The major focus will be on the computations of crystallization using nucleation theories. Two techniques, the rate equation and JMAK theory, to model the GST crystallization are investigated. The rate equation is deemed to be a more accurate method. However due to the excessive computational effort required, it is not practical to use this technique in real PCM simulation. JMAK model, on the other hand, is less accurate, but if proper parameters are used, the results can be compatible with the rate equations. It was found that the most crucial parameters used in the model is the activation energy, which is function of the heating rate. At lower heating rate, the activation energy is about 2 eV, while at higher heating rate the activation energy is around 0.81 eV. In the real PCM operation, the heating rate is more than 106 0C, therefore the the adopted activation energy should be 0.81 eV. Simulation using the JMAK theory at high heating rate with this activation energy shows good results in comparisons with the measurements available. Therefore, the JMAK model was adopted to simulate the crystallization process in a PCM device. Previously, Wang [4] has conducted similar simulations of the PCM devices. There are two major differences between the present predictions and Wang’s results. Firstly, in the RESET operation, Wang did not include the enthalpy of fusion heat in the simulation. Secondly, in the SET operation Wang adopted the predefined temperature 170 oC as the criteria to judge the completion of phase change. Instead, the present study employs the JMAK model to determine the fraction of the crystallization. In comparisons to the results of Wang, the minimum voltage required increases slightly with the heat of fusion included in the simulation, which indicates a slightly higher power is required. On the other hand, the voltage required to complete the SET operation using the JMAK model does decrease at higher BEC width comparing with the results from Wang [4].